Introduction
The brain is one of the most complicated organs in the body. With over tens of thousands of thoughts per day, the brain has a lot going on at once. To create next generation technology, scientists and engineers are creating AI that perceive just like the human brain. In UAS, scientists are working to start using these artificial neural networks to control UAV's. Using the neural network would allow for UAS systems to be completely autonomous. The remote operator would be present just to make sure that the UAV can operate and there for technical help and safety. In a study from the Department of Computer Science and Engineering at the National Institute of Technology in India, they used a Convolutional Neural Network (CNN or ConvNet) model to navigate a UAV within a building's hallway network. Within the journal, it is stated that GPS navigation can be effected by the interior of a building, causing the signal to either weaken or not exist.
Methods
As stated above, a CNN model was used for the UAV to be controlled by an artificial neural network. A ConvNet model is an algorithm that is used for deep learning to take input images and assign importance to different objects within the image to differentiate between them. The use of this computational power can allow for UAV's to be completely autonomous without the need of a direct operator.
To test out this model, the department mentioned at the National Institute of Technology used a UAV that was equipped with the CNN model. The main purpose of the network was to test and see if it was capable of being able to navigate by obstacles and travel in different directions. Figure 1 shows an image of the UAV that was used.
Other than the ConvNet system that was used, the UAV was equipped with camera that was an aid for acquiring the images that was needed for navigation of the drone. The images that are created through the camera allow for the drone to analyze the position that it is in within the hallway and compute on where it needs to go and how far.
To test out this model, the department mentioned at the National Institute of Technology used a UAV that was equipped with the CNN model. The main purpose of the network was to test and see if it was capable of being able to navigate by obstacles and travel in different directions. Figure 1 shows an image of the UAV that was used.
 |
| Figure 1: Above is the UAV that was used while it is navigating down a hallway. |
Discussion
The use of neural network models are niche. Majority of the time the types of UAV's that are used need to be controlled by a remote operator. For the small instances where the UAS is needed or requested to be autonomous, the CNN models can be used. With the way they are created, ConvNet models tend to be successful since they model the human brain which is proven to work. For example, these neural networks can be used to inspect agricultural crops for needs of water or signs of pests; since large agricultural businesses may require large plots of land, inspections can last up to hours and some instances days to fully do so. By using UAS that is fully autonomous, it speeds up the process and creates a way to do so that is not time consuming for the business owner.
Conclusion
The Convolutional Neural Network system is a great solution for complete autonomy. Equipping UAV's with the computational power to assess the general location within the geographical world allows for the computer system to function without the need of human input. The study that was mentioned above shows great potential of what the ConvNet model can be used for and the different improvements that can be made. The system they created was an infancy model that could be improved for more complex environments and maneuvers. Continuing to invest within the neural network technology will allow for more autonomous UAV's to appear within the UAS industry. Although it will not replace controlled UAV's, it will allow for the systems and situations where autonomy would be beneficial to take its course.
References
Padhy, R. P., Verma, S., Ahmad, S., Choudhury, S. K., & Sa, P. K. (2018). Deep Neural
Network for Autonomous UAV Navigation in Indoor Corridor Environments. Procedia Computer Science, 133, 643-650. doi: 10.1016/j.procs.2018.07.099
Saha, S. (2018, December 17). A Comprehensive Guide to Convolutional Neural Networks -
the ELI5 way. Retrieved April 21, 2020, from
https://towardsdatascience.com/a-comprehensive-guide-to-convolutional-neural-
networks-the-eli5-way-3bd2b1164a53
Comments
Post a Comment